Review:

"Distributed Computing:  Implementation and Management
Strategies"
R. Khanna, Ed.
Reviewed by:  Dave Crocker

Another palliative

Every few years, our fast-moving profession invents a term to
describe the next wave of technology to solve all our problems.
Things never quite live up to the promise, of course, but that
does not stop the tidal wave of media enthusiasm.  Such has been
the fate of "Distributed Computing".  Once you get past a trivial
definition, it is never quite clear what is excluded by the term,
what is accomplished by the technology, or how to go about
getting it and using it.  This book tries to remedy these
deficiencies.  Given that the topic is not subject to quick or
clever description, the book succeeds quite well.  It is an
exposition of the ideas, technologies and techniques that can be
banded together into an enterprise-wide, computational service,
and it is rich with discussion about approaches toward developing
such a service.

The most common terms for the technologies described in the book
are "middleware" and "enterprise networking".  They attempt to
enhance raw network transport services by providing additional
building-blocks for development of distributed applications and
for operation of networked platforms.  Broadly, these encompass
three nearly independent areas of service:  configuration and
query, networked operating system, and application development.
The first allows a member of the distributed environment to
obtain information about itself and other resources on the net.
The second allows multiple computers to cooperate to form an
integrated service, e.g., distributed time and remote file
access.  The last facilitates development of end-user
applications which are, themselves, distributed among various
networked systems.  With such a range of tasks to perform you
might expect "distributed computing" technology to be complicated
to deal with.  You'd be right.

Organization and style

"Distributed Computing:  Implementation and Management
Strategies" is a series of independently-written expository
chapters.  Each is serious and relatively information-dense.
Don't look for clever asides or outrageous myth-busting
diatribes.  Many of the chapters are written by principals in the
development or promotion of the respective technologies, so
skeptical analysis is not generally present.  This does not mean
that reality is absent.  In a chapter on "Migration Strategies",
Bob Morgan cautions:  "DCE's complexity is a challenge in itself,
however.  It is likely that only organizations that are already
hip-deep in the distributed computing swamp will be able to
appreciate the benefits..."

After an introductory chapter, the book is divided into 3 parts
and a series of Appendices which really are short chapters
forming a bonus section.  In Part 1 six chapters discuss basic
technologies, including the much-heralded DCE, as well as the
ONC+ enhancements to the much-used ONC/NFS.  Jeff Schiller's
chapter on Distributed System Security is as good an introduction
to the topic of network-based security as I've seen.  Security
discussions tend to sound like black-magic incantations to the
non-expert, yet Jeff covers the ground thoroughly and clearly.
(And you probably won't miss much if you skip the few math
notations.)

The second Part of the book comprises 5 case studies, in keeping
with the book's effort to be pragmatic.  Included are
descriptions of CMU's Andrew and MIT's Athena projects, as well
as discussion of "institutional" implementation and use projects
at the University of Michigan, Hewlett-Packard, and Eastman
Kodak.  All of these chapters are useful; the Andrew and Athena
ones are particularly excellent for the perspectives they offer.

Part 3 discusses "strategies" and "issues".  In other words,
these chapters try to step back and discuss things using a broad
perspective.  The first of these is a downright clever chapter in
which Khanna got vendor strategy statements from Sun, HP, IBM and
Microsoft (with one of the appendices coming from Apple).  Page-
flipping among these consecutive summaries was particularly
interesting for comparing their architectural charts.  If you
have any expectation that this realm of technology will make your
life simpler, the charts should finish that off.

No quick answers

Media discussion about distributed computing has heavily
emphasized OSF's work on DCE and DME.  Both are covered in detail
in the book, with Morgan's chapter discussing adoption and use.
As long as you remember that these real experiences were by
highly skilled staff in a highly advanced environment, the
discussion adds flesh to the otherwise-spare repertoire of
knowledge about DCE realities.

Distributed computing technology provides an enhanced
infrastructure, rather than direct solution to specific end-user
requirements.  This books does the same thing.  It provides
extensive detail about an area of new technology and service and
about the approaches that can be taken to incorporate the
capabilities.  It does not contain any cookbooks or procedural
checklists.  It provides a basis for doing extended analysis and
a discussion of the current industry experiences.  Like the move
to distributed computing, reading the book will be an investment
in the future.   If you develop distributed systems or must make
acquisition decisions for them, I recommend the investment.